Over the past few months, it seems that Google engineers have mentioned query refinement an awful lot. First it was in the site search box within the SERPS, which didn’t go over very well once people noticed their site search results were surrounded by ads for competing sites. Then it was Danny Sullivan’s interview with Marissa Mayer where she mentioned that previous query refinement is making it’s way into the organic results. More recently Danny mentioned it again in the Search 4.0 article.
As somebody who likes to watch the search engines and figure out a general idea of why Google ranks pages the way they do this is interesting to me. It seems to be coming more and more prevalent in the SERPS these days than ever before. This makes sense to me, as search engines can gauge their relevancy by how few query revisions it takes for a searcher to get to their destination. If it takes me 3 revisions to find what I’m looking for on MSN, and only 2 or 1 to get it on Yahoo, I could conclude that for that query Yahoo provides more relevant results, to me.
By providing site links and site search boxes, Google is doing the best it can to help you limit the number of revisions you have to make before you find a relevant page. Over time this will get better and better as they collect more data and see common occurrences of certain query revisions. Bill Slawski has pointed out that search engines have access to query logs which could include:
- The positioning of URLs in the search results,
- A URL the searcher clicked on after viewing the search results,
- The number of times the searcher clicked on the URL,
- The time the search query was received,
- The search query received before the present query, or;
- Any other type of information relating to the search query or search results desired by the search engine.
Using this data the engines will be getting a much clearer signal of intent behind queries, how queries may be semantically related, and ultimately how to get the user to their destination in less time.
Site Links Based on Query Refinement?
I started thinking about this today when I was questioned by a colleague as to why a site in the SERPS had a site search box and how we could go about replicating that. Then I stumbled upon the official entry from Google about the topic and was surprised at how forthright the Google reps were regarding how they’re triggered.
“..there were lots of searchers who would type the name of a specific website as if they wanted to teleport, but would then immediately issue another more a refined search within this site.
This feature will now occur when we detect a high probability that a user wants more refined search results within a specific site.”
Many may disagree with this, but I think that a Google site search box is basically a case of having site links on steroids, and Google’s intention on including them in the SERPs is the same reason they have sitelinks, which is to get people to their destination ASAP.
I don’t know of a way to determine queries before or after a query is entered (MSN has a tool which allows you to do this, but I couldn’t glean anything from it), but I did find some similarities between the navigation paths from the home page and site links that appeared. The most visited pages people clicked through to off the homepage weren’t exactly the same pages that appeared in the sitelinks, but I did notice a positive correlation (5 of the 7 site links).
I can file this one away under things that I have 0 proof of but that make sense. If even 20% of users go from the SERPS to the homepage to another page within the site, that page is well worth including in the Site Links. It’s almost the same as if instead of clicking through off the SERPS to the homepage, they did one more query refinement in Google to get to the page they ultimately wanted. I’d also like to note that there were still 2 that I have no idea how they got there. The sites I was looking at all had Google Analytics installed, which makes it that much easier for Google to see how people click through the site.
Now I’m aware that these findings aren’t entirely related to query refinement, and I’ve gone off on a mild tangent these last few paragraphs, but it’s late and I’m a little rusty after having not posted in a month.
Has anybody else figured out anything conclusive further related to site links and how to get them?